Question on experimental aspect of Mass Spectrometry

[abhishekdub...@gmail.com]

Dear All

Lately I have been trying to get more and more understanding of Mass Spectrometry and associated data analysis.

Currently, I am probing the following problem : We performed a pull-down experiment alongwith proper controls to identify band on SDS-PAGE that is probably ligand for our protein of interest. Now, we want to identify the ligand. The species is Mus Musculus.

Based on the availability of resources, I can only perform a MALDI based MS and MS/MS. The instrument that we have is Waters Synapt G2-Si.

Following is how the data is recorded :

1. The sample is spotted on MALDI plate and then we take a MS1 scan. We select a peak and then fragment it further.

2. The MS/MS file contains data from multiple scans that happen over the same spot. Each scan corresponds to one spectrum in the data file. One can say that the data is recorded in time, therefore we have 50-100 scans at the same spot.

Even if we fix the CID energy, still some of the scans are high-intensity while some are low-intensity and/or noisy.

Following are my questions :

1. Is it ok to pick the good-quality, high-intensity scans from the MS/MS data of each peak ?

2. I had a MS/MS data of a particular MS1 peak in which only 5 scans out of 100 are of good quality. If I run tide-search and percolator with just these 5 scans, I don't get a significant result. Also, percolator gives warning of too small sample read. Does it makes sense to do it this way ?

Looking forward to get some help

Best regards

Abhishek

[William S Noble]

Hi Ahbishek,

My attempted answers are interleaved below.

On Sun, Jan 10, 2021 at 3:06 AM abhishekdub...@gmail.com <abhishekdub...@gmail.com> wrote:

Dear All

Lately I have been trying to get more and more understanding of Mass Spectrometry and associated data analysis.

Currently, I am probing the following problem : We performed a pull-down experiment alongwith proper controls to identify band on SDS-PAGE that is probably ligand for our protein of interest. Now, we want to identify the ligand. The species is Mus Musculus.

Based on the availability of resources, I can only perform a MALDI based MS and MS/MS. The instrument that we have is Waters Synapt G2-Si.

Following is how the data is recorded :

1. The sample is spotted on MALDI plate and then we take a MS1 scan. We select a peak and then fragment it further.

2. The MS/MS file contains data from multiple scans that happen over the same spot. Each scan corresponds to one spectrum in the data file. One can say that the data is recorded in time, therefore we have 50-100 scans at the same spot.

Even if we fix the CID energy, still some of the scans are high-intensity while some are low-intensity and/or noisy.

Following are my questions :

1. Is it ok to pick the good-quality, high-intensity scans from the MS/MS data of each peak ?

I am guessing that by "ok" you mean in a statistical sense.  I think the answer is "yes," as long as your assessment of quality is based only on the MS/MS data and does not make use of the FASTA file of proteins.

 

2. I had a MS/MS data of a particular MS1 peak in which only 5 scans out of 100 are of good quality. If I run tide-search and percolator with just these 5 scans, I don't get a significant result. Also, percolator gives warning of too small sample read. Does it makes sense to do it this way ?

Running Percolator with just 5 scans will not work.  Percolator needs to train a classifier using hundreds of scans, at a minimum.  If you have other data that you can use to train Percolator, you can save the trained model and then use that model to score your 5 scans of interest. See this paper for details:

https://pubs.acs.org/doi/10.1021/acs.jproteome.9b00780

You can also do this easily using our new implementation of Percolator in mokapot:

https://www.biorxiv.org/content/10.1101/2020.12.01.407270v1

However, this strategy will not yield statistical confidence estimates, again because you only have five scans.  To get a q-value from any method (Percolator or otherwise), you need a large number of scans to generate decoy scores. Sorry!

Bill

 

Looking forward to get some help

Best regards

Abhishek

--
You received this message because you are subscribed to the Google Groups "crux-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to crux-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/crux-users/5bdd3e5b-28ca-4219-a58f-c368d5fd3cf9n%40googlegroups.com.